DISCUSSION PAPER SERIES
IZA DP No. 4062
The Impact of Child Labor and School Quality on Academic Achievement in Brazil Márcio Eduardo G. Bezerra Ana Lúcia Kassouf Mary Arends-Kuenning March 2009
Forschungsinstitut zur Zukunft der Arbeit Institute for the Study of Labor
The Impact of Child Labor and School Quality on Academic Achievement in Brazil Márcio Eduardo G. Bezerra Universidade de Sao Paulo
Ana Lúcia Kassouf Universidade de Sao Paulo
Mary Arends-Kuenning University of Illinois at Urbana-Champaign and IZA
Discussion Paper No. 4062 March 2009 IZA P.O. Box 7240 53072 Bonn Germany Phone: +49-228-3894-0 Fax: +49-228-3894-180 E-mail:
[email protected]
Any opinions expressed here are those of the author(s) and not those of IZA. Research published in this series may include views on policy, but the institute itself takes no institutional policy positions. The Institute for the Study of Labor (IZA) in Bonn is a local and virtual international research center and a place of communication between science, politics and business. IZA is an independent nonprofit organization supported by Deutsche Post Foundation. The center is associated with the University of Bonn and offers a stimulating research environment through its international network, workshops and conferences, data service, project support, research visits and doctoral program. IZA engages in (i) original and internationally competitive research in all fields of labor economics, (ii) development of policy concepts, and (iii) dissemination of research results and concepts to the interested public. IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion. Citation of such a paper should account for its provisional character. A revised version may be available directly from the author.
IZA Discussion Paper No. 4062 March 2009
ABSTRACT The Impact of Child Labor and School Quality on Academic Achievement in Brazil* We analyze the impact of child labor on school achievement using Brazilian school achievement test data from the 2003 Sistema Nacional de Avaliação da Educação Básica (SAEB). We control for the endogeneity of child labor using instrumental variable techniques, where the instrumental variable is the average wage for unskilled male labor in the state. Using our preferred OLS estimates, we find that child labor causes a loss in students’ school achievement. Children and adolescents who do not work have better school performance than students who work. Up to two hours of work per day do not have a statistically significant effect on school performance, but additional hours decrease student’s achievement. Differences in work conditions affect school performance. For high school students in Portuguese, compared to students who have schooling as their only activity, students who work only at home score 4 percent lower on the tests. Those students who only work outside the house are worse off than those who only work within the house, with test scores decreasing by 5 percent. Students who work both inside and outside the house have the lowest test scores of all the working conditions, decreasing by up to 7 percent.
JEL Classification: Keywords:
I21, J13, J22, O15
child labor, school achievement, Brazil
Corresponding author: Mary Arends-Kuenning Department of Agricultural and Consumer Economics University of Illinois at Urbana-Champaign 408 Mumford Hall MC-710 1301 W. Gregory Urbana, IL 61801 USA E-mail:
[email protected]
*
Paper prepared for the seminar on the Quality of Education in Latin America, Universidad Iberoamericana, Mexico City, February 2007. An earlier version of this paper was presented at the Population Association of America annual meetings, Los Angeles, CA, April 2006. We thank Susan Parker for her helpful comments.
1. Introduction In recent years, Brazil has experienced an impressive decline in child labor. According to the national household survey, the Pesquisa Nacional de Amostra por Domicilios (PNAD), in 1992, about 15 percent of Brazilian children aged 5 to 15 worked, compared to 8 percent in 2005 (IBGE 2005). With respect to educational indicators such as illiteracy rates and years of schooling, Brazil still lags behind other Latin American countries. However, during the 1990s, school attendance increased, principally in primary school and for students aged 7 to 14. In 1992, 87 percent of the children aged 7 to 14 attended school. By 2005, this percentage reached 97 percent (IPEA 2005). A possible reason why Brazil continues to lag other countries in school achievement despite the increases in school attendance is that a high percentage of students work while they attend school. According to the 2003 PNAD data, of Brazilian children aged 7 to 15, 88.1 percent only study, 1.0 percent work and do not study, 8.4 percent combine work with study, and 2.6 percent neither work nor study. 1 This statistic shows that there are a significant number of children and adolescents who continue to divide their time between working and studying, which could harm their school achievement. This study contributes to the literature by analyzing the direct impact of child labor on the academic progress of students as measured by standardized achievement tests. Authors such as Gunnarsson et al. (2004), Psacharopoulos (1997), Heady (2003), Akabayashi e Psacharopoulos (1999), Stinebrickner and Stinebrickner (2003) studied the effect of early child labor on student achievement test scores. However, this study differs 1
In the 2003 PNAD, the rural populations of the states of Rondônia, Acre, Amazonas, Roraima Pará and Amapá in the Northern region of Brazil are not included.
2
from previous studies because we investigate how the number of hours worked by young people might harm student learning. In addition, we find that student performance is affected differently by work conducted inside the household than by work in the labor market. Estimating the relationship between child labor and schooling is complicated because students who work might do poorly in school, but poor performance in school can also lead to families deciding that children should invest more time in work. School characteristics, family characteristics, and individual characteristics all affect both child labor and school achievement. We address the issue of causality through using the instrumental variable of low-skilled wages, which affect child labor without directly affecting school achievement. We control for school characteristics using a rich set of variables available from a school census. A novel aspect of this paper is the use of controls for individual students’ motivation in the regression analyses. 2. Data The Ministry of Education (MEC) through INEP makes available educational statistics to measure the quality of the Brazilian primary school system. The Sistema Nacional de Avaliação da Educação Básica (SAEB) (National system of basic education evaluation) was started in 1990 and since 1995 has administered a national standardized test every two years. The data set used in this study is from SAEB, (INEP, 2003), which makes available microdata that include standardized test scores in Portuguese and Mathematics for students in the 4th and 8th grades of ensino fundamental (primary school) and the third year of ensino médio (secondary school) in public and private schools. The SAEB data
3
set includes information about students’ study habits, students’ sociocultural characteristics, the characteristics and teaching practices of teachers and school administrators, managerial mechanisms, and school infrastructure. The SAEB is collected from a representative sample of the population of students who are enrolled in the three grades of interest in the schools chosen to participate in the School Census. 2 This sample is stratified by the following criteria: grade, state of residence, whether the school is public or private, location of school (state capital city, large cities defined as greater than 200,000 habitants, small cities), and size of school. In 2003, approximately 300,000 students, 17,000 teachers, and 6,000 administrators in 6,270 schools participated in the SAEB, covering all 26 states plus the Federal District (INEP/SAEB 2005). In this study, we include the data for eighth grade and the third year of high school, because few fourth graders work. Data are only available for students who live in urban areas because rural schools are not included in the sample for these higher-grade levels. SAEB administered standardized, multiple-choice exams designed to measure students’ abilities and capacities in Portuguese (with a focus on reading comprehension) and Mathematics. 3 The exam scores are mapped into cumulative performance scales, implying that students who are placed at a given level are competent at the skills required at the previous levels of the scale. Based on percentage scales, SAEB classifies students
2
The School Census is a national and annual undertaking, covering basic education at different levels (preschool, primary, and secondary schooling) and types (regular, special, and youth and adult education). It provides a data base with educational information about all the school establishments involved in basic education, both public and private. 3 More detailed descriptions about the learning competencies and abilities can be found in the research paper SAEB in INEP (2002).
4
into five levels of achievement in Portuguese and in Mathematics—very critical, critical, intermediate, adequate, and advanced. 3. The econometric model To analyze the effect of child labor on student school achievement test scores, we use two estimation methods—ordinary least squares and two-stage least squares. In addition to analyzing whether the student works or not, the analyses also consider the number of hours worked and the work conditions, specifically whether the student only works, studies and works only at home, studies and works only outside the home, or studies and works in both locations. The hours of work and work conditions have important impacts on students’ achievement because they determine how much time students can spend on school activities. To obtain an estimate of the impact of child labor on school performance, control variables are included such as the individual characteristics of the student and of the family environment, characteristics of the school including the quality of school infrastructure, and the supply of educational services, geographic characteristics of the schools, and, uniquely, controls for the student’s motivation to study. These factors have been shown to affect both child labor and school achievement (Barros and Mendonca 1996, Psacharopoulos 1997, Cavalieri 2000, Gunnarsson et al., 2004, Soares 2002). The dependent variables are the Portuguese and Mathematics test scores, which are continuous variables, for each grade that is evaluated. The test score can vary, theoretically, from 0 to 500 points. However, according to INEP, no students obtained the scores at the limits of the interval (values of 0 and 500 points). Therefore, the use of OLS rather than a tobit model is appropriate.
5
The model to measure the effect of early labor on the students’ school performance is as follows:
Di = α + βLi + δX i + ρM i + θE s + λG s + ε i ,
i = 1,...., n
, s = 1,...., S
(9)
Where D is the dependent variable that represents the school achievement test score of each student i in Mathematics or Portuguese in the 8th grade of primary school and the third year of secondary school. The exogenous variables are represented by Li, which refers to the information about child labor for each student i, Xi, which are the individual and family characteristics of students, Mi, which are the variables that proxy for students’ motivation to study, E, which represents the schools’ infrastructure and educational resources for each school, s, and G, which are the geographic aspects of each school s. In addition, there is an error term ε, which represents the unobservable characteristics that affect learning achievement and that are not captured by the specified variables. The primary purpose of this work is to identify the effect of child labor on school achievement, obtained with the estimated parameter β . We expect that ∂D